Comments on: Running your own hardware Vs EC2 and RightScale — Part 2 http://www.derivante.com/2008/09/16/running-your-own-hardware-vs-ec2-and-rightscale-part-2/ to obtain or receive from a source Sat, 30 Jul 2011 08:51:15 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: earthgecko http://www.derivante.com/2008/09/16/running-your-own-hardware-vs-ec2-and-rightscale-part-2/comment-page-1/#comment-418 earthgecko Mon, 15 Mar 2010 13:01:49 +0000 http://justinleider.com/?p=40#comment-418 Nice to see some bad experiences with the cloud. Unfortunately everyone seems to think they are bulletproof. Researching methods on which to build fully redundant cloud infrastructure is always interesting. Unfortunately, the hype about EC2 or any other cloud or grid type infrastrucutre does not take into account the fine details. Everyone (developers and CTOs) just think slapping it all on the cloud is great. Things like EBS/XFS kernel module bugs that do not allow instances to remount EBS volumes after rebooting never get taken into account. Then digging to determine what DOESN'T have problems, becames a spiralling, looping journey through Amazon forums and Google groups. Making it all fully redundant in the cloud is quite complex, the 24 page slideshows do not go into that. Complex and costly... as you add this functionality and that functionality, branching here and there. Building in the cloud is not the problem..... recovering in the clouds way it goes wrong is... because of some kernel module or likewise which is somewhat out of the controller of the maintainer. Anyway, I just wanted to say thanks for sharing your experiences. PS - EBS and XFS? Check your kernel.. dmesg | head -n 1 If it is Linux version 2.6.21.7-2.fc8xen-ec2-v1.0 then perhaps start looking through some of the following threads as it seems it may just happen again.. even with EBS (not certain in the EBS boot scenario). http://developer.amazonwebservices.com/connect/thread.jspa?threadID=28968 http://developer.amazonwebservices.com/connect/message.jspa?messageID=151961 http://groups.google.com/group/ec2ubuntu/browse_thread/thread/0cfca179e77a880f?pli=1 The latest Ubuntu kernel has the XFS module compiled into it so those problems should not occur on it. I find it hard to get my mind around running production appliances on EC2, when EC2 itself is a single point of failure. Is there a storm brewing in the clouds..? Nice to see some bad experiences with the cloud. Unfortunately everyone seems to think they are bulletproof. Researching methods on which to build fully redundant cloud infrastructure is always interesting.

Unfortunately, the hype about EC2 or any other cloud or grid type infrastrucutre does not take into account the fine details. Everyone (developers and CTOs) just think slapping it all on the cloud is great.

Things like EBS/XFS kernel module bugs that do not allow instances to remount EBS volumes after rebooting never get taken into account. Then digging to determine what DOESN’T have problems, becames a spiralling, looping journey through Amazon forums and Google groups.

Making it all fully redundant in the cloud is quite complex, the 24 page slideshows do not go into that. Complex and costly… as you add this functionality and that functionality, branching here and there.

Building in the cloud is not the problem….. recovering in the clouds way it goes wrong is… because of some kernel module or likewise which is somewhat out of the controller of the maintainer. Anyway, I just wanted to say thanks for sharing your experiences.

PS – EBS and XFS? Check your kernel..
dmesg | head -n 1

If it is Linux version 2.6.21.7-2.fc8xen-ec2-v1.0 then perhaps start looking through some of the following threads as it seems it may just happen again.. even with EBS (not certain in the EBS boot scenario).

http://developer.amazonwebservices.com/connect/thread.jspa?threadID=28968
http://developer.amazonwebservices.com/connect/message.jspa?messageID=151961
http://groups.google.com/group/ec2ubuntu/browse_thread/thread/0cfca179e77a880f?pli=1

The latest Ubuntu kernel has the XFS module compiled into it so those problems should not occur on it.

I find it hard to get my mind around running production appliances on EC2, when EC2 itself is a single point of failure.

Is there a storm brewing in the clouds..?

]]> By: Randy Rizun http://www.derivante.com/2008/09/16/running-your-own-hardware-vs-ec2-and-rightscale-part-2/comment-page-1/#comment-74 Randy Rizun Wed, 08 Oct 2008 01:18:47 +0000 http://justinleider.com/?p=40#comment-74 Hi Justin- just a thought- http://www.subcloud.com might be a better solution instead of an ec2 instance + nfs, in terms of both cost and reliability (i.e, no need to run an ec2 instance for nfs) Hi Justin- just a thought- http://www.subcloud.com might be a better solution instead of an ec2 instance + nfs, in terms of both cost and reliability (i.e, no need to run an ec2 instance for nfs)

]]>